WebWright: built for action, not just browsing

Tracking since May 18, 2026.

Changelog

• Jun 6, 2026
  description

  WebWright is an autonomous AI agent that lives in your browser sidebar — not a chatbot, but a true agentic AI that perceives web pages, reasons about them, and takes real actions on your behalf. It's a lightweight AI browser extension for hands-free web automation, turning plain-English goals into completed, multi-step tasks.
  Agent Mode — type a goal in natural language and watch the AI agent navigate sites, click buttons, fill forms, and complete complex multi-step tasks, with a real-time action log of every step it takes.
  Chat Mode — ask questions about any page you're viewing, with optional vision support so the model can read layouts, charts, and on-page images directly.
  Research Mode — enter a topic and receive a polished, multi-source HTML research report in minutes — scraped, summarized, and synthesized from across the open web.
  Workflows — record any repetitive web task once and replay it with a single click for one-tap browser automation.
  Personal Info Vault — a local-only store of your details that powers instant, secure AI form filling. Your data never leaves your device.
  Built for reliability, it uses 4-tier vision escalation (DOM analysis → screenshot with Set-of-Marks overlays → expanded markers → raw coordinate clicks) and anti-loop detection, so it recovers and finishes the job where other AI browsing agents stall.
  Total model freedom — supports 8 LLM providers out of the box with full model flexibility. Bring your own API key, or run entirely on a local LLM for fully private, zero-cloud AI browsing.
  Privacy-first by design. The extension is under 5 MB, has zero dependencies, loads no remote code, makes no analytics or telemetry calls, and runs only when you open it. All data is stored locally in your browser.

  WebWright — Your Free Autonomous AI Agent for the Web — not a chatbot, but a real agent that perceives web pages, reasons about them, and takes actions on your behalf.
  🤖 Agent Mode
  Type a goal in plain English and watch it work. WebWright navigates sites, clicks buttons, fills forms, and completes multi-step tasks — all with a real-time action log so you see exactly what it's doing.
  💬 Chat Mode
  Ask questions about any page you're viewing. Optional vision support lets the model see layouts and charts, not just text.
  🔎 Research Mode
  Enter a topic and get a polished, multi-source HTML report in minutes.
  ⚡ Workflows
  Record repetitive tasks once and replay them with a single click.
  🔐 Personal Info Vault
  Stores your details locally for instant form filling — your data never leaves your device.
  Built to go where other tools get stuck:
  ✓ Automatic 4-tier vision escalation: DOM analysis → screenshot with Set-of-Marks overlays → expanded markers → raw coordinate clicks
  ✓ Anti-loop detection that recovers instead of freezing
  ✓ 8 LLM providers out of the box with full model flexibility — bring your own API key or run entirely local for zero-cloud privacy
  Private and lightweight by design:
  ✓ Under 5 MB, zero dependencies
  ✓ Loads no remote code
  ✓ Makes no analytics or telemetry calls
  ✓ Runs only when you open it
  ✓ All data stored locally in your browser — nothing is sent to any developer-controlled server

• May 24, 2026
  description

  WebWright is an autonomous AI agent that lives in your browser sidebar — not a chatbot, but a real agent that perceives web pages, reasons about them, and takes actions on your behalf. In Agent Mode, type a goal in plain English and watch it navigate sites, click buttons, fill forms, and complete multi-step tasks with a real-time action log. In Chat Mode, ask questions about any page you're viewing, with optional vision support so the model can see layouts and charts. In Research Mode, enter a topic and get a polished multi-source HTML report in minutes. Record repetitive tasks as Workflows and replay them with one click. A built-in Personal Info Vault stores your details locally for instant form filling — your data never leaves your device. WebWright features automatic 4-tier vision escalation (DOM analysis → screenshot with Set-of-Marks overlays → expanded markers → raw coordinate clicks) and anti-loop detection so it recovers where other tools get stuck. Supports 8 LLM providers out of the box with full model flexibility — bring your own API key or run entirely local for zero-cloud privacy. The extension is under 5 MB, has zero dependencies, loads no remote code, makes no analytics or telemetry calls, and runs only when you open it. All data is stored locally in your browser; nothing is sent to any developer-controlled server.

  WebWright is an autonomous AI agent that lives in your browser sidebar — not a chatbot, but a true agentic AI that perceives web pages, reasons about them, and takes real actions on your behalf. It's a lightweight AI browser extension for hands-free web automation, turning plain-English goals into completed, multi-step tasks.
  Agent Mode — type a goal in natural language and watch the AI agent navigate sites, click buttons, fill forms, and complete complex multi-step tasks, with a real-time action log of every step it takes.
  Chat Mode — ask questions about any page you're viewing, with optional vision support so the model can read layouts, charts, and on-page images directly.
  Research Mode — enter a topic and receive a polished, multi-source HTML research report in minutes — scraped, summarized, and synthesized from across the open web.
  Workflows — record any repetitive web task once and replay it with a single click for one-tap browser automation.
  Personal Info Vault — a local-only store of your details that powers instant, secure AI form filling. Your data never leaves your device.
  Built for reliability, it uses 4-tier vision escalation (DOM analysis → screenshot with Set-of-Marks overlays → expanded markers → raw coordinate clicks) and anti-loop detection, so it recovers and finishes the job where other AI browsing agents stall.
  Total model freedom — supports 8 LLM providers out of the box with full model flexibility. Bring your own API key, or run entirely on a local LLM for fully private, zero-cloud AI browsing.
  Privacy-first by design. The extension is under 5 MB, has zero dependencies, loads no remote code, makes no analytics or telemetry calls, and runs only when you open it. All data is stored locally in your browser.

Permissions & access

Permissions

activeTabscriptingsidePanelstoragetabsdebuggerwebNavigation

Host access

<all_urls>

Screenshots